Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
Subject: Re: [RFC 1/1] s390/pci: expose UID checking state in sysfs
To:     Bjorn Helgaas <helgaas@kernel.org>,
        Christian Brauner <christian.brauner@ubuntu.com>
Cc:     linux-pci@vger.kernel.org, linux-kernel@vger.kernel.org,
        linux-s390@vger.kernel.org, Pierre Morel <pmorel@linux.ibm.com>,
        Peter Oberparleiter <oberpar@linux.ibm.com>,
        Viktor Mihajlovski <mihajlov@linux.ibm.com>
References: <20210113185500.GA1918216@bjorn-Precision-5520>
From:   Niklas Schnelle <schnelle@linux.ibm.com>
Message-ID: <675aa466-59ea-cf8a-6eec-caa6478ba4cd@linux.ibm.com>
Date:   Thu, 14 Jan 2021 14:20:10 +0100
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:78.0) Gecko/20100101
 Thunderbird/78.5.0
MIME-Version: 1.0
In-Reply-To: <20210113185500.GA1918216@bjorn-Precision-5520>
Content-Type: text/plain; charset=utf-8
Content-Language: en-US
Content-Transfer-Encoding: 7bit
Precedence: bulk


On 1/13/21 7:55 PM, Bjorn Helgaas wrote:
> On Wed, Jan 13, 2021 at 08:47:58AM +0100, Niklas Schnelle wrote:
>> On 1/12/21 10:50 PM, Bjorn Helgaas wrote:
>>> On Mon, Jan 11, 2021 at 10:38:57AM +0100, Niklas Schnelle wrote:
>>>> We use the UID of a zPCI adapter, or the UID of the function zero if
>>>> there are multiple functions in an adapter, as PCI domain if and only if
>>>> UID Checking is turned on.
>>>> Otherwise we automatically generate domains as devices appear.
>>>>
>>>> The state of UID Checking is thus essential to know if the PCI domain
>>>> will be stable, yet currently there is no way to access this information
>>>> from userspace.
>>>> So let's solve this by showing the state of UID checking as a sysfs
>>>> attribute in /sys/bus/pci/uid_checking
> 
>>>> +/* Global zPCI attributes */
>>>> +static ssize_t uid_checking_show(struct kobject *kobj,
>>>> +				 struct kobj_attribute *attr, char *buf)
>>>> +{
>>>> +	return sprintf(buf, "%i\n", zpci_unique_uid);
>>>> +}
>>>> +
>>>> +static struct kobj_attribute sys_zpci_uid_checking_attr =
>>>> +	__ATTR(uid_checking, 0444, uid_checking_show, NULL);
>>>
>>> Use DEVICE_ATTR_RO instead of __ATTR.
>>
>> It's my understanding that DEVICE_ATTR_* is only for
>> per device attributes. This one is global for the entire
>> Z PCI. I just tried with BUS_ATTR_RO instead
>> and that works but only if I put the attribute at
>> /sys/bus/pci/uid_checking instead of with a zpci
>> subfolder. This path would work for us too, we
>> currently don't have any other global attributes
>> that we are planning to expose but those could of
>> course come up in the future.
> 
> Ah, I missed the fact that this is a kobj_attribute, not a
> device_attribute.  Maybe KERNEL_ATTR_RO()?  Very few uses so far, but
> seems like it might fit?
> 
> Bjorn
> 

KERNEL_ATTR_* is currently not exported in any header. After
adding it to include/linuc/sysfs.h it indeed works perfectly.
Adding Christian Brauner as suggested by get_maintainers for
their opinion. I'm of course willing to provide a patch
for that move should it be desired.

@Bjorn apart from the correct macro do you have a preference
for either suggested path /sys/bus/pci/zpci/uid_checking vs
/sys/bus/pci/uid_checking?

For completeness some further internal discussion lets
us tend to rather name it to "unique_uids" but I guess that
doesn't make a difference to non s390 people ;-)